AI and jobs in logistics and food: which tasks are changing?

Will AI make jobs disappear? What the research shows so far

The International AI Safety Report 2026, published on 3 February 2026, gives a sober summary of the state of the evidence. General-purpose AI will probably automate a wide range of cognitive tasks, especially in knowledge work. The early evidence, however, shows no effect on overall employment, though it does show some signs that demand for early-career workers is falling in some occupations that are highly exposed to AI, such as writing. According to the report, economists disagree about the size of future effects: some expect lost jobs to be offset by new ones, others that broad automation could significantly depress employment and wages.

The OECD painted a similar picture in its Employment Outlook of July 2023. In 2022, it surveyed more than 2,000 employers and 5,300 workers in manufacturing and the financial sector in seven OECD countries. The conclusion: the effect of AI on the number of jobs had been limited up to that point. Many early adopters preferred not to make staff redundant and instead adjusted their workforce through slower hiring, voluntary departures and retirement. Some companies said that AI could in fact help with an ageing workforce and staff shortages. For now, according to the OECD, AI is changing the jobs themselves and the skills they require rather than replacing jobs.

As early as January 2023, CPB wrote that technological change in the Netherlands cuts both ways: automation causes a considerable decline in a number of existing jobs and tasks, but new technology and growing prosperity also create new jobs and tasks. That is no comfort to anyone who sees their own task disappear. It is, however, a reason to have the conversation at task level rather than at the level of whole roles.

Exposure is not the same as replacement

Many headlines about job losses can be traced back to what are known as exposure indices: estimates of the share of tasks in an occupation that AI could technically perform. The ILO published a refined version in May 2025. Worldwide, one in four workers is in an occupation with some exposure to generative AI; in high-income countries, the figure is 34 percent. The highest category, where most of the tasks are exposed, accounts for 3.3 percent of global employment. Administrative occupations are the most exposed. The ILO's conclusion: because human input remains necessary, most jobs will change rather than become redundant.

In a follow-up brief of February 2026, the ILO puts it even more sharply: exposure indicators show technical susceptibility, not labour market outcomes. They do not measure whether automation is cost-effective, how work processes change or how employment and wages adjust. According to the ILO, they therefore cannot be read as a prediction of job losses.

For the Netherlands, the first indications are mixed. Three Rabobank economists wrote in ESB (a Dutch economics journal) on 22 January 2026 that in the occupational groups most susceptible to generative AI, mainly occupations with many graduates of higher professional education (hbo) and university (wo), the number of open vacancies in the second quarter of 2025 was almost 25 percent lower than at the end of 2022. In the other 102 occupational groups, the fall was less than 10 percent. In the third quarter of 2025, more than 13 percent fewer young people aged 15 to 25 were working in those susceptible occupations than at the end of 2022, while employment for young people in other occupations rose by 3 percent. The authors themselves write that it is too early to say whether and how generative AI is structurally changing the labour market. Another ESB article, of 8 July 2026, found no broad decline in employment among recent graduates, except among graduates of university business administration programmes. It also points to explanations given by other researchers, such as macroeconomic uncertainty and the rise of working from home.

What you can take from this for logistics and food: in so far as there is an effect, it is mainly in office work, and the first signals concern early-career workers. Whether that pattern also occurs in your sector has not yet been measured. It is, however, a reason to keep an eye on the places where the trade is learned: the junior planner and the new employee in the back office.

How far has AI come in transport and manufacturing?

Less far than the debate suggests. According to CBS, one in six companies used AI in 2025, double the figure of two years earlier. In the transport and storage sector that was 7 percent, in manufacturing 16 percent. Among companies with 50 to 250 persons employed it stood at 45 percent, among those with 250 or more at 66 percent. Companies use AI most often for marketing or sales (35 percent) and for business administration or management tasks (32 percent). Only 4 percent of the companies using AI used it for logistics purposes.

Between 15 September and 12 December 2025, UWV surveyed 3,550 employers, and on 10 March 2026 it wrote that AI can lead to certain tasks disappearing, but that new jobs are also being created. Many occupations are changing because of AI, and training and development are therefore becoming more important.

Which tasks are changing in logistics and food?

A job is a bundle of tasks. AI can prepare or take over some tasks in that bundle, and not others. Below, role by role, is where the change lies. We deliberately give no percentages: how much of this AI can prepare in your company can only be established with your own baseline measurement.

• Planner (transport, production or cultivation): AI calculates alternative plans, proposes a new plan when something drops out and supplies a demand forecast. Building the plan becomes less of a manual job; assessing, choosing and handling exceptions become more important. Driving and rest times, customer agreements and the choice itself stay with the planner. How that works out when drivers are in short supply is set out in our piece on the driver shortage and AI.
• Administration and back office: reading and keying in orders, packing slips, consignment notes, CMRs and invoices is typical of the administrative work that the ILO sees as most exposed. There is less retyping; checking, resolving exceptions and contact with the customer or haulier take up more of the day.
• Customer service: AI can prepare draft replies, a summary of a file and answers to standard questions about delivery times or shipments. The ILO cites customer service roles as an example of roles exposed to generative AI. The conversation over a complaint, during a recall or with a difficult customer remains human work.
• Quality department: AI can support record-keeping, deviation reports and gathering documents for an audit, and image recognition on the line acts as an extra pair of eyes. The assessment, the release and the signature on a HACCP record stay with the quality officer. The OECD describes an aerospace company that had new turbine blades for jet engines checked with AI-assisted visual inspection. Before that, inspectors sat for long periods in a darkened room and examined the blades through a magnifying glass; according to the OECD, their working environment improved.
• Warehouse and production floor: generative AI affects physical work least directly. The ILO places manual occupations at the edge of the network of occupations, where the knock-on effect of AI is smaller. Here, change is more likely to come from software for pick routes and task allocation and from robotisation, as we describe in our piece on AI in the warehouse. Something does change for team leaders: they must be able to judge what a proposal from the system is based on.

AI and the tight labour market: relief, not replacement

The labour market is less tight than in 2022, but still tight. According to CBS, there were 95 open vacancies per 100 unemployed people in the second quarter of 2026, compared with 142 at the peak in the second quarter of 2022. In transport and storage, 14,400 vacancies were open, and in manufacturing 29,400. In the first quarter of 2026, UWV classed 87 of the 93 occupational groups as tight or very tight.

For truck drivers, the labour market is not easing, according to UWV. About 6,000 to 7,000 vacancies are open, about 4 in 10 drivers are aged 55 or over, and the Researchcentrum voor Onderwijs en Arbeidsmarkt (ROA, the Research Centre for Education and the Labour Market) expects the tight market to last until 2030. UWV also mentions the self-driving truck: according to the Dutch government, it can be on the road for the first time by 2029 at the latest, and within ten years do about 10 percent of the work of ordinary trucks. Even then, according to UWV, drivers remain important for checking the load, customer contact and safety.

In the food industry, producers cited a shortage of workers (27 percent) as the main obstacle to growth, ABN AMRO reported in December 2025; we set out what that means for automation in our piece on the labour shortage in the food industry. More broadly, according to UWV, 58 percent of employers were affected by the tight labour market in 2025, and 52 percent of employers with vacancies are more often hiring people who still need to be trained.

It is striking that UWV rates the tightness for administrative occupations such as secretaries and bookkeeping staff as average, while the shortages in technical occupations and in health and social care are large. In such a market, the question is rarely who you can do without. The question is how the people you have can spend less time on retyping and searching, and have more time left for the work where they make the difference.

What does AI ask of your employees?

According to the UWV survey, about half of employers think that employees need different training because of AI; of those employers, 70 percent offer training or courses on AI. Of the employers that use AI, about 60 percent give their employees guidance on it; some 40 percent do not. About half of employers expect skills such as collaboration, social and communication skills, digital, critical and ethical skills and adaptability to become more important.

The OECD found that outcomes are better where workers have been trained to work with AI and where its introduction has been discussed with them. Nor is this optional in legal terms. Article 4 of the EU AI Act on AI literacy has applied since 2 February 2025. Following the amendment by the Digital Omnibus, in force since 27 July 2026, it asks organisations that deploy AI to take measures to support the AI literacy of their staff, taking into account knowledge, experience, education and the context in which the system is used. A specific level for each employee does not have to be guaranteed. What that requires in practice is set out in our piece on AI literacy.

At AIBA, our starting point is that the human decides and signs. A planner who can weigh up a proposal from the system, a quality officer who can assess a signal from the camera and a back-office employee who spots a misread document: those are the skills that matter when tasks shift.

What to do now as an employer

Waiting is not a neutral choice, because the use of AI grows even without a policy. Six steps give you a firm footing without your promising anything you cannot deliver:

• Carry out a task analysis for each role. For the planner, the back office, the quality department and the team leader, put the main tasks on paper, noting for each task the time it takes, whether it is repetitive work, what AI could prepare and who decides and signs. Start with the administrative tasks, where exposure is greatest.
• Choose one work process and record a baseline measurement. How many hours, how many errors, how many exceptions. Only then can you show later what has changed, including to your employees.
• Train by role, not in general. A planner needs different knowledge from a quality officer. Record who has been trained in what.
• Involve the works council early. A works council is mandatory in an enterprise where at least 50 people normally work. Under Article 25(1)(k) of the Wet op de ondernemingsraden (Dutch Works Councils Act), the works council has the right to advise on a proposed decision to introduce or change an important technological provision. Its consent is required for rules on staff training and for rules on measures that are aimed at or suitable for monitoring or checking the attendance, behaviour or performance of employees (Article 27(1)(f) and (l)).
• Pay attention to early-career workers. If AI prepares the simple tasks through which juniors learn the trade, you need to plan deliberately how they will still gain that experience, for example by having them check the system's proposals with an experienced colleague alongside.
• Keep an eye on the SLIM scheme (the Dutch subsidy scheme for learning and development in SMEs). SMEs can apply for a subsidy for a company review with a training plan or for introducing a learning method, up to €24,999 per application (agricultural businesses €20,000), for 60 percent of the eligible costs. Training costs themselves are excluded. The 2026 application windows have closed; the last one closed on 7 September 2026, and the next window has not yet been announced. The current status is on our SLIM page (in Dutch).